top of page

From Neural Responses to Population Behavior

The essay next is a cut from the paper: From Neural Responses to Population Behavior: Neural Focus Group Predicts Population-Level Media Effects. Which was written by Emily B. Falk, Elliot T. Berkman, and Matthew D. Lieberman at University of California.

Can neural responses of a small group of individuals predict the behavior of large-scale populations? In this investigation, brain activations were recorded while smokers viewed three different television campaigns promoting the National Cancer Institute’s telephone hotline to help smokers quit (1-800-QUIT-NOW). The smokers also provided self-report predictions of the campaigns’ relative effectiveness. Population measures of the success of each campaign were computed by comparing call volume to 1-800-QUIT-NOW in the month before and the month after the launch of each campaign. This approach allowed us to directly compare the predictive value of self-reports with neural predictors of message effectiveness. Neural activity in a medial prefrontal region of interest, previously associated with individual behavior change, predicted the population response, whereas self-report judgments did not. This finding suggests a novel way of connecting neural signals to population responses that has not been previously demonstrated and provides information that may be difficult to obtain otherwise.

Can small groups of individuals efficiently predict population-level behavior? People are notoriously limited in their ability to predict their own future behavior and accurately identify their internal mental states through verbal and written selfreports (Nisbett & Wilson, 1977). Furthermore, explicitly asking participants to reflect on such internal mental states (e.g., “Why do you like this?”) has been shown to alter the outcome and quality of judgments (Wilson & Schooler, 1991). Thus, it is not surprising that public-health media messages selected using traditional focus groups—which rely on these forms of self-report—are also imperfect predictors of population-level responses (Noar, 2006).

Recent research has identified neural indicators of individuals’ future behavior that may be inaccessible using self-reports (Berns & Moore, 2012; Brewer, Worhunsky, Carroll, Rounsaville, & Potenza, 2008; Falk, Berkman, Mann, Harrison, & Lieberman, 2010; Knutson, Rick, Wimmer, Prelec, & Loewenstein, 2007; Kosten et al., 2006; Paulus, Tapert, & Schuckit, 2005; Tusche, Bode, & Haynes, 2010). However, it has not been previously demonstrated whether neural responses to persuasive messages in a small group of individuals also forecast behavioral responses at the population level (e.g., in a city or state).

To examine this question, we partnered with public health organizations that had produced television ads designed to help smokers quit. We used ads from three campaigns in a functional MRI (fMRI) investigation conducted in a separate location from where the ads were aired. Participants in our study (smokers who intended to quit) viewed ads from each campaign while their neural activity was measured. In a previous study, we used the same task and sample to demonstrate that overall neural activity across all the ads predicted individual smoking reduction in the month following the scan, above and beyond the participants’ self-reports of intention to quit, quitting-related self-efficacy, and their ability to relate to the ads (Falk, Berkman, Whalen, & Lieberman, 2011). In the analyses reported here, we used those data together with new data (about population-level outcomes) to answer an orthogonal question: Would neural activity in response to the different ad campaigns predict the effectiveness of the campaigns among a larger group of new individuals? To address this question, we used the fMRI data and self-report predictions of the ads’ effectiveness to rank the campaigns. We then compared these rankings with the actual population-level success of the campaigns. Neither this analysis, the brain-activationbased and self-report measures of the ads reported in this article, nor the population data were reported in the previous study. The approach described here is novel because it directly links neural responses with behavioral responses to the ads at the population level.

All three measures of participants’ self-reported projections of ad effectiveness produced the same mean ranking of the ad campaigns (Table 2): Campaign B was ranked highest, followed by Campaign A, and then Campaign C (Fig. 1b). Industry experts who were familiar with the campaigns also ranked Campaigns B and A above C. In contrast with the self-report measures, the prediction based on the participants’ mean neural activity in the MPFC ROI during ad exposure suggested a different campaign order: C > B > A (Table 2; Fig. 1c).

Given that there are six possible ways to order the three campaigns, each ordering has a 1/6 probability of occurring by chance. Therefore, in addition to examining group means, we also examined the frequency with which each ordering occurred across subjects (Fig. 2). Consistent with the mean ratings, our results showed that 33% of the individual rankings based on MPFC activity suggested the order C > B > A. A chisquare test confirmed that the proportion of C > B > A orderings suggested by MPFC activation was significantly above chance, χ2(1, N = 30) = 5.97, p = .015, whereas no other ordering of MPFC data appeared above chance level (16.67%). This result also indicates that C > B > A was selected more frequently than any other order, providing an unambiguous prediction from MPFC activity. The proportion of self-report rankings mirrored the ordering suggested by mean self-report ratings across self-report metrics (see Fig. S2 in the Supplemental Material for results of each self-report metric), which suggests a different, unambiguous prediction (B > A > C) from self-report data. In other words, MPFC and self-report metrics each produced clear but discrepant predictions of the population-level response.

At the population level, each of the ad campaigns led to increases in call volume to the National Cancer Institute’s Smoking Quitline, ranging from 2.8- to 32-fold increases (Table 2; Fig. 1d) compared with the month prior to the launch of each campaign. Increases in call volume to the Quitline in the month after the campaigns were launched were taken as a proxy for the population-level success of each campaign. The ordering of population-level success (based on call-volume increase) was C > B > A, which was consistent with the neural predictions (C > B > A) but different from the self-report predictions (B > A > C). This ordering remained the same both before and after adjusting for a variety of potential differences between media markets, including media weight purchased, time of year, unemployment rate, smoking rate, and tobacco-control policies.

Thus, both the average and most frequently observed neural responses in our MPFC ROI correctly ordered the success of the ad groups at the population level, whereas self-reports of our participants and anecdotal evaluations of industry experts did not. To confirm the reliability of this result, we examined the distances between individual MPFC rankings and the modal (correct) ordering using a distance-based metric for ranked data, weighted Kendall’s tau. To the degree that individual MPFC rankings consistently favored one prediction (in this case, selecting the best ad campaigns), the average distance between observed individual rankings and the modal response should be smaller than the distance between rankings obtained by chance and any modal ranking. Results of this analysis supported the hypothesis that MPFC activations provided a more consistent ranking of the best ads than what would be expected by chance: τw = .3667, mean expected τw = .5, t(29) = −2.0708, p = .0474 (or, given the strong directional nature of our hypothesis, p = .0237, one-tailed).

51 visualizações0 comentário

Posts recentes

Ver tudo

Doença de Alzheimer

Por João Caetano "Somos nossa memória, somos esse quimérico museu de formas inconstantes, esse montão de espelhos rompidos.” Jorge Luis Borges, escritor buenairense considerado um dos autores mais inf

bottom of page